NIM error creating mksysb_resource via NIM 2

[pdtak]

Has anyone run into issues creating a mksysb_resource via NIM?
We get the following messages

0042-001 nim: processing error encountered on "master":
0042-006 m_mkbosi: (From_Master) connect A remote host refused an
attempted connect operation.

This happens on the LPARs on the same frame between master and client.

Any ideas?

 

[khalidaaa]

Possible causes:

1. Check for any firewall blocking ports between the master and client.
2. Are both your servers (Master and Client) on the same subnet? If not then they have to be.
3. Name resolution! Have you checked nslookup and ping from both servers?
On master & client (all should get resolved to correct IP addr. you configured for NIM)

#host master
#host masters ip
#host client
#host client ip

4. If all of the above are ok then try removing the machine definition on the master and redefine it again:

Please do following to reset the client

At Master

nim -Fo reset <client>
nim -o deallocate -a subclass=all <client>
smitty nim_rmmac >> <client>

At Client

smitty niminit >> <client> <ent> <master>

Then try the creating the mksysb resource again.

Regards,
Khalid

 

[pdtak]

Thanks for your answer Khalidaaa, but the problem was bigger than what you and I thought.
We have an Etherchannel setup on our P5-590's, and here's the answer from IBM support:

Unforunately, Etherchannel (and other multiple adapter protocols,
like VIPA, Virtual IP Addressing) is not supported with NIM.

NIM requires, when performing a network boot for a client, that
-all- network traffic occur between only -two- adapters: The NIM
primary network adapter and the client's configured network adapter
(for boot purposes).

Etherchannel attempts to "load balance" network traffic by distributing
the network traffic for the system among the multiple adapters it has
in its configuration. This means that a BOOTP request can come in
one of the Etherchannel network adapters, but the response to the
client goes out another adapter, and is ignored by the client,
because the response packet was send by a different adapter than
the adapter that received it.
This behaviour prevents NIM from performing network boots of clients.

The only recourse is to either remove the Etherchannel configuration
from the NIM master system, or separate a -dedicated- network
adapter for use by NIM -only-, such that only NIM traffic can go out
and be received by this adapter.

There is also an unsupported method of getting Etherchannel and
NIM to cooperate. You can set the Etherchannel device to
"netif_backup", with one adapter acting as the 'primary' network
adapter for the system, and the other second adapter acting as a
"Backup Adapter", so that all network traffic goes in and out of
the 'primary' adapter, and the 'backup' adapter is only on standby,
to take over if the 'primary' adapter fails. This method is
unsupported, because the Etherchannel failover to the backup
adapter could happen in the middle of a NIM network boot of a
client, and will cause the network boot to fail.

 

[khalidaaa]

Wow! As you said, this is unpredictable and its my first time to know about!

So you really deserve a star for updating the problem!

Regards,
Khalid

 

[pdtak]

Khalid,
IBM support person said that there are many PMR's associated with this issue.
We are trying to re-configure our network to make it work, but it's not going to be an easy task.
Thanks.
Paul

 

[MoreFeo]

I usually allways configure etherchannels with one main adapter and the other as a backup adapter.
With older versions of AIX creating an etherchannel with 2 main adapters could make some trouble, and even with AIX 5.3 it can make some trouble depending on the ethernet switches.